| name | research-cataloging |
| description | Router for research, archiving, and cataloging: provenance bundles, OCR/transcription, exhaustive source sweeps, translation promotion, catalog reconciliation, NocoDB/Airtable ops, and archival image compression. Use before conversion skills to cover a whole site/course/source set, or whenever you need a transcript. Read one reference. Triggers: Wayback, OCR, coverage ledger, captions, yt-dlp, Whisper. |
Research & Cataloging (router)
Sources/gaps: references/sources.md maps the research/cataloging references; references/coverage-gaps.md parks source completeness, transcript, database, and handoff standards.
Read exactly one reference file below; each holds the full discipline for its area (deeper worked procedures, helper scripts, and agent manifests live in a same-named subdirectory next to it).
- references/archival-research.md — per-subject research bundles: web source trails with snapshots + attributed imagery, OCR/transcription of scanned PDFs into layered corpora, venue/exhibition/artwork dossiers.
- references/source-sweep.md — exhaustive coverage of bounded sources: inventory every node, coverage ledger, open every page, never judge from titles.
- references/catalog-reconciliation-research.md — making structured records internally consistent: object identities, evidence tiers, image provenance/coverage audits, archive paths, downstream records.
- references/source-translation-workflow.md — translation workflows: promoting translations into catalog records, quote fields, bilingual sources.
- references/nocodb-catalog-management.md — NocoDB operations: bases/tables/fields, records, linked-record relations, attachments, gallery views, CSV sync.
- references/image-archival.md — shrinking oversized images (PNG/TIFF/PSD) to visually identical smaller files: WebP/JPEG/AVIF re-encode, verify lossless, never overwrite originals. (Applies outside research contexts too — any "huge image, compress it" request.)
- references/openai-transcription-chunking.md — transcribing audio/video into auditable transcript evidence: caption-first preflight with
yt-dlp (YouTube, Instagram, Reels, TikTok, Vimeo, podcasts), then OpenAI/Whisper ASR when captions are missing, with chunking, hallucination audits, and per-chunk retries. Use whenever you need the actual words from a recording — including studying a creator's voice — rather than a webpage's paraphrase.
- references/coverage-gaps.md — parked candidate standards for pagination completeness, transcript evidence, database write posture, source-to-skill handoff packets, and scanned-corpus sidecars.
Chains: research → reconciliation → NocoDB sync is the common pipeline; references/archival-research.md is the entry reference and its Related-references block routes onward.
Also here: references/airtable-mcp.md (Airtable MCP setup/auth/troubleshooting in Codex + record ops).
